Research by neuroscientists at the University of Chicago shows how short-term, working memory uses networks of neurons differently depending on the complexity of the task at hand.
The researchers used modern artificial intelligence (AI) techniques to train computational neural networks to solve a range of complex behavioral tasks that required storing information in short term memory. The AI networks were based on the biological structure of the brain and revealed two distinct processes involved in short-term memory. One, a “silent” process where the brain stores short-term memories without ongoing neural activity, and a second, more active process where circuits of neurons fire continuously.
The study, led by Nicholas Masse, Ph.D., a senior scientist at UChicago, and senior author David Freedman, Ph.D., professor of neurobiology, was published this week in Nature Neuroscience.
While it might be nice to use a $4,000 oscilloscope in a lab at a university or well-funded corporate environment, a good portion of us won’t have access to that kind of equipment in our own home shops. There are a few ways of getting a working oscilloscope without breaking the bank, though. One option is to find old CRT-based unit for maybe $50 on craigslist which might still have 60% of its original 1970s-era equipment still operational. A more reliable, and similarly-priced, way of getting an oscilloscope is to just convert a device you already have.
The EspoTek Labrador is an open-source way of converting a Raspberry Pi, Android device, or even a regular run-of-the-mill computer into a working oscilloscope. It’s a small USB device with about a two square inch PCB footprint that includes some other features as well like a signal generator and logic analyzer. It’s based on an ATxmega which is your standard Arduino-style AVR microcontroller but geared for low power usage. It looks as though it is pretty simple to use as well, and the only requirements are that you can install the software needed for the device on whatever computing platform you decide to use.
While the Labrador is available for sale at their website, it is definitely a bonus when companies offer products like this but also release the hardware and software as open source. That’s certainly a good way to get our attention, at least. You can build your own if you’d like, but if you’d rather save the time you have pre-built options. And it doesn’t hurt that most of the reviews of this product seem to be very favorable (although we haven’t tried one out ourselves). If you’d prefer an option without a company backing it, though, we have you covered there too.
A group of researchers at Sandia National Laboratories have developed a tool that can cross-train standard convolutional neural networks (CNN) to a spiking neural model that can be used on neuromorphic processors. The researchers claim that the conversion will enable deep learning applications to take advantage of the much better energy efficiency of neuromorphic hardware, which are designed to mimic the way the biological neurons work.
The tool, known as Whetstone, works by adjusting artificial neuron behavior during the training phase to only activate when it reaches an appropriate threshold. As a result, neuron activation become a binary choice – either it spikes or it doesn’t. By doing so, Whetstone converts an artificial neural network into a spiking neural network. The tool does this by using an incremental “sharpening process” (hence Whetstone) through each network layer until the activation becomes discrete.
According to Whetstone researcher Brad Aimone, this discrete activation greatly minimizes communication costs between the layers, and thus energy consumption, but with only minimal loss of accuracy. “We continue to be impressed that without dramatically changing what the networks look like, we can get very close to a standard neural net [in accuracy],” he says. “We’re usually within a percent or so on performance.”
Neuromorphic systems carry out robust and efficient neural computation using hardware implementations that operate in physical time. Typically they are event- or data-driven, they employ low-power, massively parallel hybrid analog/digital VLSI circuits, and they operate using the same physics of computation used by the nervous system. Although there are several forums for presenting research achievements in neuromorphic engineering, none are exclusively dedicated to this increasingly large research community. Either because they are dedicated to single disciplines, such as electrical engineering or computer science, or because they serve research communities which focus on analogous areas (such as biomedical engineering or computational neuroscience), but with fundamentally different goals and objectives. The mission of Neuromorphic Engineering is to provide a publication medium dedicated exclusively and specifically to this field. Topics covered by this publication include: Analog and hybrid analog/digital electronic circuits for implementing neural processes, such as conductances, neurons, synapses, plasticity mechanisms, photoreceptors, cochleae, etc. Neuromorphic circuits and systems for implementing real-time event-based neural processing architectures. Hardware models of neural and sensorimotor processing systems, such as selective attention systems, coordinate transformation systems, auditory and/or visual processing systems, sensory fusion systems, etc. Implementations of neural computational systems found in insects, birds, mammals, etc. Embedded neuromorphic systems, including actuated or robotic platforms which process sensory signals and interact with the environment using event-based sensors and circuits. To ensure high quality and state-of-the-art material, publications should demonstrate experimental results, using physical implementations of neuromorphic systems, and possibly show the links between the artificial system and the neural/biological one they model.
Artificial life (often abbreviated ALife or A-Life) is a field of study wherein researchers examine systems related to natural life, its processes, and its evolution, through the use of simulations with computer models, robotics, and biochemistry.
A new photonic chip design drastically reduces energy needed to compute with light, with simulations suggesting it could run optical neural networks 10 million times more efficiently than its electrical counterparts.
Image: courtesy of the researchers, edited by MIT News.
The most essential Pentagon suppliers will be the ones that master robotics and artificial intelligence.
Look across the Potomac River toward Rosslyn, where the corporate logos of government contractors crown a parade of office towers that follows the river past the Pentagon. The skyline, like America’s defense industrial landscape, is changing. Soon, 25,000 Amazon employees will be climbing the Metro escalators to work in Crystal City each morning along with the tens of thousands of workers from military, intelligence, and the defense industry organizations.
The arrival of Amazon’s HQ2 in the cradle of U.S. government contracting comes at a portentous time for the Defense Department. Technology is altering what makes us strong, prosperous, and secure. The defense industrial base is becoming the strategic innovation base. Today’s leading digital companies have disrupted every industry they have touched, from publishing to automotive. Could Amazon and the rest of the “FAANG companies”—Facebook, Apple, Netflix, and Google—or one of a handful of pure-play artificial-intelligence companies, such as the authors’ SparkCognition, become fixtures of this new industrial base?
